{"id":25813389,"url":"https://github.com/mkgeiger/neural-network","last_synced_at":"2026-06-14T10:34:13.068Z","repository":{"id":279101283,"uuid":"937719623","full_name":"mkgeiger/neural-network","owner":"mkgeiger","description":"Simple neural-network supporting most importent features like Convolutional-/Fully Connected network completely written in Ansi C","archived":false,"fork":false,"pushed_at":"2025-02-23T18:43:39.000Z","size":67,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-02-23T19:35:01.229Z","etag":null,"topics":["adagrad","adam-optimizer","categorical-cross-entropy","convolutional-neural-network","fully-connected-network","glorot-uniform","image-classification","leaky-relu","mean-square-error","momentum-optimization-algorithm","neural-network","relu-layer","rmsprop-optimizer","sgd-optimizer","sigmoid","softmax-layer","softsign","tanh"],"latest_commit_sha":null,"homepage":"","language":"C","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/mkgeiger.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2025-02-23T18:31:30.000Z","updated_at":"2025-02-23T18:51:07.000Z","dependencies_parsed_at":"2025-02-23T19:45:06.666Z","dependency_job_id":null,"html_url":"https://github.com/mkgeiger/neural-network","commit_stats":null,"previous_names":["mkgeiger/neural-network"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/mkgeiger/neural-network","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkgeiger%2Fneural-network","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkgeiger%2Fneural-network/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkgeiger%2Fneural-network/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkgeiger%2Fneural-network/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/mkgeiger","download_url":"https://codeload.github.com/mkgeiger/neural-network/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkgeiger%2Fneural-network/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34318523,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-14T02:00:07.365Z","response_time":62,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["adagrad","adam-optimizer","categorical-cross-entropy","convolutional-neural-network","fully-connected-network","glorot-uniform","image-classification","leaky-relu","mean-square-error","momentum-optimization-algorithm","neural-network","relu-layer","rmsprop-optimizer","sgd-optimizer","sigmoid","softmax-layer","softsign","tanh"],"created_at":"2025-02-28T02:37:55.531Z","updated_at":"2026-06-14T10:34:13.052Z","avatar_url":"https://github.com/mkgeiger.png","language":"C","funding_links":[],"categories":[],"sub_categories":[],"readme":"# neural-network\nA lightweight neural network library written in ANSI-C supporting prediction and backpropagation for Convolutional- (CNN) and Fully Connected  (FCN) neural networks.\n\n# Features\n\n## Different layer types:\n- `Input layer`: 1-dimensional layer used e.g. as input for a FCN.\n- `Dense layer`: 1-dimensional fully connected hidden layer of a FCN, having weights and an activation function.\n- `Dropout layer`: 1-dimensional hidden layer of a FCN to reduce overfitting by randomly deactivating some nodes.\n- `Output layer`: 1-dimensional layer which represents the outputs of a FCN, having weights and an activation function.\n- `Input2D layer`: 2-dimensional layer used e.g. as input for a CCN.\n- `Conv2D layer`: 2-dimensional layer of a CNN, having for each produced output channel (feature map) one 3D-kernel (3D-filter).\n  For now only filter rows, filter columns, filter numbers and filter stride is supported. The activation function type is limitted to only ReLU and Leaky ReLU.\n- `Max Pooling2D layer`: 2-dimensional layer of a CNN, downsampling the input by taking the maximum value over an input window.\n  For now only the most common window size of 2x2 with a stride of 2 is supported.\n- `Avr Pooling2D layer`: 2-dimensional layer of a CNN, downsampling the input by taking the average value over an input window.\n  For now only the most common window size of 2x2 with a stride of 2 is supported.\n- `Flatten2D layer`: 2-dimensional layer of a CNN. This layer is used to make the transition from a CNN (2-dimensional) to a FCN (1-dimensional).\n\n## Different activation functions:\n- `Sigmoid`: s-shaped curve, only for Dense- and Output layers.\n- `ReLU`: Rectified Linear Unit curve, for Dense-, Output- and Conv2d layers.\n- `Leaky ReLU`: Leaky Rectified Linear Unit curve, for Dense-, Output- and Conv2d layers.\n- `Tanh`: hyperbolic tangent curve, only for Dense- and Output layers.\n- `Softsign`: softsign curve, only for Dense- and Output layers.\n- `Softmax`: softmax function, only for the Output layer.\n\n## Different loss functions:\n- `MSE`: mean squared error\n- `Categorical cross entropy`: softmax loss (used always in combination with the Softmax activation in Output layers)\n\n## Different random function types (used to initialize weights, biases and filter values when training begins):\n- `Uniform` (uniform distributed): general purpose, lightweight models\n- `Normal` (gaussian distributed): general purpose\n- `Glorot uniform`: general purpose, sigmoid, tanh, ReLU activations, classification\n- `Glorot normal`: sensitive training, sigmoid, tanh or softmax activations, regression\n- `HE uniform`: wide Layers, ReLU and variants (Leaky ReLU), classification\n- `HE normal`: very deep networks, ReLU and variants (Leaky ReLU), regression\n\n## Different optimizer function types (used to adjust weights, biases and filter values during the training):\n- `SGD`: stochastic gradient descent\n- `SGD with decay`: stochastic gradient descent with decay\n- `Adapt`: stochastic gradient descent with adapt\n- `Momentum`: stochastic gradient descent with momentum\n- `RMSProp`: root mean square propagation\n- `Adagrad`: adaptive gradient algorithm\n- `Adam`: adaptive moment estimation\n  \n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkgeiger%2Fneural-network","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmkgeiger%2Fneural-network","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkgeiger%2Fneural-network/lists"}